1 Mb/s and Higher Data Rate PHYIMAC: GFSK and … 1993 Doc: IEEE P8Q2,ll-931l38 IEEE 101.11 WIRELESS...

Septeniber 1993 Doc: IEEE P8Q2,ll-931l38

IEEE 101.11

WIRELESS ACCESS METHODS AND PHYSICAL LAYER SPECIFICA nONS

Tide: 1 Mb/s and Higher Data Rate PHYIMAC: GFSK and FQPSK

Author: Kamilo FEHER Department of Electrical and Computer Engineering

University of Califomia, Davis, Davis CA 9S616 and Digcom Consulting Group, Inc.

ABSTRACf

Preamble and PHY header bits for the rccendy standardized 1 Mbls rate GFSKFH system are suggested. Bits to develop" ... mea"' for negotiating a switch to bigher data rates from the data rate defined ... " are also proposed. .

For higher data rate FH. OS sp~ad spectrum. and infrared applications, we propose FQPSK and FQPSK-KF nonlinearly amplified (NLA) robust power and spectrally efficient solutions. These wireless systems operated with simplest threshold detectors in an interference environment (binary robust eye diagrams in the I and Q channels of the coherent demodulator) have an increased bit rate of 1.5 Mbls while the 4-state systems have a bit rate capability of over 2 Mbls (with 3 Mbls potential), in a 1 MHz bandwidth. Here bandwidth is defined at 99% or -20 dB power spectral density (PSD), that is, the "conservative" FCC defmition is used. An integrated AC1-FCC intcrpRtalion. suggested in this submission, would lead to significantly higher data rates. Productlhardware measu~ment ~sults, computer analysis and study demonstrate the following key parameters:

1 Mbls

19.3 dB (1S.S*)

1.1 Mb/s

10.5 dB

-with more complex receiver ~ proceaar

1.5 Mbls 2.048 Mbls+ 3 Mbls +

15.7 dB IS dB 19.8 dB(?)

The BER perfonnance of the proposed FQPSK modems in a NLA-A WON (Additive White Gius-$i.1ln Noise) and also Rayleigh faded interference cona-olled environment [BER = f(CII)] is evaluated. A performance comparison with other proposed uhigher bit rate systems," is undertaken. Based on the critical BER ::I f(Eb/No) and highest . possible bit rate criteria. the FQPSK.KF family of NLA radio/modems ourperform 4-FSK and FLOQAM systems proposed in References [IS; 24]. Notice of patent applicability has been given in Document IEEE P.802.11·93/138 [5].

Low power, miniaturized cost efficient VLSI based FQPSK and FQPSK-KF modem/radios, suitable for smallest PCMCIA card implementations are presented in an accompanying paper presented by Dr. S. Kato of NIT [7]. These coherent systems have a rapid synchronization time, shon preamble, and reduced number of PHY header bit requirements.

Preliminary data indicate ~at reduced data file transfer time/message delay and increased throughput could be attained with more robust modems havinl lower EblNo and CJI requirements. A detailed investigation is proposed.

Submission pale I It FdieI'. uc Divis .

September 1993 Doc; IEEE P802.11-93/138

1. INTRODUCTION

During the Denver, July 12-15, 1993, IEEE 802.11 meetings, a motion in regards to the IEEE standardization of the Frequency Hopped (FH)-Wireless Local Area Network (WLAN) was approved. Reference Document: [IEEE P.802.11-93/117, Denver, July 1993, pp. 7-10]. This standard requires that

FH·PHYs shall be capable of operating using GFSK with BT = O.S and a minimum deviation of 160 kHz with a data rate of 1 Mb/s.

In this submission the unedited draft wording of the standard is used as the exact content of the final edited version is not known to me. It was stated that the FCC 15.247 rules apply for this 2.4 GHz to 2.48 GHz ISM band application and that the maximal RF 99% bandwidth is 1 MHz. The reader is advised to consult the official standardization documents. The following was also approved:

A means for negotiating a switch to higher data rates from the data rate defined above is for further study.

This second part of the approved standard, "a means of negotiating a switch to higher data rates ... " has been followed up by the newly formed "IEEE P.802.11 WLAN Higher Data Rate FH-PHY Ad-Hoc Interest Group" with co-chairs W. Moyers and N. Silberman. More than 30 Ad Hoc members have expressed a substantial interest in the investigation and possible standardization of higher (higher than 1 Mb/s) rate FH-PHY systems.

In this report, Nonlinearly Amplified (NLA) techniques which could attain a considerably higher data rate than the standardized basic rate of 1 Mb/s are described. NLA techniques have been recommended due to their increased RF power!battery efficiency ratio. Proposed simple and robust coherent binary filtered offset QPSK (or FQPSK) systems attain about 1.5 Mb/s while with 4-1evel (in I and Q) more than 2 Mb/s could be transmitted. With more elaborate 8·8 states up to 3 Mb/s could be transmitted in 1 MHz. Preamble and PHY header bits for the standardized GFSK and for "switch to higher data rates" are suggested.

This presentation and paper is a continuation of our IEEE 802.11, July 13, 1993, Denver, CO paper, Ref. [6].

Submission page 2 K. Feher, UC Davis

September 1993 Doc; IEEE PS02.11-93/138

2. NOMENCLATURE AND ABBREVIATIONS

Frequently used terms in this report and also in other submissions to the IEEE 802.11 Physical Layer (PHY) Committee are listed below.

FQPSK Feher's patented fllter/processor for QPSK (tenn "FQPSK"), offset QAM and other applications Ref. [1-5].

The term "FQPSK" represents a general term for a large class of processed filtered QPSK, offset QPSK, offset QAM (OQAM) and other systems. Illustrative examples presented in this report.

FQPSK -1 One of the basic Intersymbol-Interference Free (IJF) based patented "filter" processors [1]. Other modified systems are claimed in [1-6].

FQPSK-KF Kata-Feher based invention a cross correlation between the I and Q quadrature data streams. Applies to FQPSK and others. S. Kato of NIT Japan and K. Feher, UC Davis [2; 13; 17]

FQPSK-4*4 Four signaling states in the demodulated I and Q signals. The transmitter is a nonlinearly amplified (NlA) offset 16 QAM or 16-OQAM having 4*4 independent NLA signal states.

GFSK

GMSK

NLA

Gaussian Frequency Shift Keying (this modem has a variable modulation index and is suitable for noncoherent demodulation)

Gaussian Minimum Shift Keying (MSK) has a modulation index of m = 0.5. For exact values of m = 0.5 coherent as well as noncoherent demodulation can be used

nonlinearly amplified

3. FQPSK AND FQPSK.KF NONLINEARLY AMPLIFIED (NLA) RADIO

A description of FQPSK-NLA modems and radio systems has been presented to the IEEE 802.11 committee during the Denver, CO, July 13, 1993 meeting [6]. A detailed description has been presented in more than 30 IEEE and other journal and conference papers, patents [1-5], and a book [13]. For details, please see the reference list in [6].

In Fig. 1 and Fig. 2, a generic block diagram of FQPSK modulation and parts of radio systems which could be suitable for

FH Frequency Hopped Spread Spectrum

OS Direct Sequence including COMA

IR Infrared-PHY

WLAN applications are illustrated. The FQPSK family of NLA modems/radios has many other possible applications not described in these basic conceptual block diagrams and/or report of FH-PHY.

Submission page 3 K. Feher, ac Davis


FQPSK.KF: Correlated Signal Processor· A Review of Basic Concepts

The Kato and Feher (KF) modem radio or "FQPSK-KF' technique [2; 13; 17; 7] relates to a signal processor invention which is particularly useful for bandwidth efficient and power efficient nonlinearly amplified (NLA) wireless and other radio system applications. In order to reduce the envelope fluctuations of an UF (Intersymbol and Jitter Free) processor or filter, or of a similar related processor, [see reference patents 1-5], a simple "crosscorrelator" is introduced. Dlustrative circuits and concepts of the basic first set of filters, called "UF' have been described in numerous patents and papers including [1-7; 10-13; 17; 25]. The resultant output of the UF encoder is illustrated in Fig.4a. A typical crosscorrelated output signal at the I-input drive of a quadrature mixer is illustrated in Fig. 4c [2; 7; 17].

One of the simplest "filtered" QPSK is the "FQPSK-l" set of signals. This set obtained by deleting the crosscorrelator. A large class of modulated crosscorrelated signals can be generated by numerous independent combinations of the processed baseband wave form or IJF and the coefficients/algorithms of the correlator. The basic FQPSK invention [Patent Ref. "Filter" 1; 3; 4] describes the generation of a very large class of useful baseband signals. Combined with the crosscorrelator or "X" functions, and followed by a simple DSP (or other technique) low-pass filtering or processing, a large class of two-level, as well as multilevel signals can be generated. These are all suitable for nonlinear amplification. The crosscorrelator patent could also be used for processing non "UF' signals and non "FQPSK" signals.

Block DiagramslCircuits/Results

In Fig. 3, an FQPSK.KF transmitter/receiver is illustrated. The FQPSK-KF (KatolFeher correlated FQPSK) has a very simple Baseband Processor (BBP) in an offset QPSK or equivalent filtered offset QAM or for short FOQAM [13] configuration. The "F" filter provides one of the many possible simple processed baseband bandlimited waves, as explained in Patents [1-5] and also in Ref. [6; 7; 13; 17]. Following a basic crosscorrelation or "X", simple digital (or analog) LPF (low pass filters) are inserted for final spectral shaping. For example, these LPF's could be chosen from a large class of 4th or 8th order filters having Butterworth, Chebycheff, Bessel or approximately Double Jump or variable shape responses. The receiver is a conventional offset mode QPSK receiver, such as described by NIT's Dr. Kato during this September 1993 Atlanta 802.11 meeting [6; 17].

One of the many possible Intersymbol-Jitter Free (IJF) waveforms of the FQPSK baseband signal is illustrated in Fig. 4(a). Other types of non UP signals can also be generated by the inventions of [1-5]. The block diagram of a crosscorrelated or FQPSK-KF modulator and one illustrative resulting "I" and "Q" baseband drive signal of the quadrature modulation (OQPSK or offset QAM, i.e., OQAM) inputs is illustrated in Fig. 4(b) and Fig. 4(c).

A very simple circuit and implementation diagram concept is shown in Fig. 5. In this circuit discrete Ie components are used. The "simple digital filter" could be a 4th or 8th order tuter as described previously. The schematic diagram for a possible DSP implementation is shown in the lower part of Fig. 5 [25]. Large scale or VLSI single chip implementation are low-cost, low-power and simple [7; Dr. Kato et. at, NTI].


c:

EL en fa o ::s

'1::3 c;

; ..,..

r Tr~iI Pin 01 OSPaup-F-QPSK - - - - - 1 I

r -I

fit _

300kWalD 3 ...

BBP 1(1)

(J/ F-QPSK

BBP oIf.QPSK

Q(I)

MAC Frequency --... eo.oa

1

I'

I

L

~Hoppc:r

VCX~FH-SS 'JDNA

2AQIb •..

1:

POWQ

Conuol MAC

o.a. Qua from 1.1

Fig. 1 Conceptual diagram of a transmitter of a FH or ns spread spectrum radio system. Bit rate "fb" could be replaced by "chip rate fc.•• A fairly general Baseband Procei&Ql' (HBP) chip based on nsp implementation of a class of FQPS~ Nonlinearly Amplified (NLA) systems is ilIusttatcd

Amc/Ula

V,) s::: 0-

Et en en .... , o =

"0 ~

~

~I • 1

611

ax

Frequenc) Synabesiza

Fig. 2

r

1-

1

RP-IUCbip

,---

vao At .. 0.. ..

- - - -811"'-111'

Ilf - OS. - Oup ... d"'1/1

a...-IShopna

Receiver with one IF stage for coherent demodulation of conventional offset QPSK (OQPSK), offset QAM (OQAM) and of FQPSK or FQPSK-KF

September 1993 Doc: IEEE P802.11-93/138

Fig. 3

submission

Data

BPP

SIP

IIQ Dem.

F Filtel' LPF

X

IIQ Mod.

F Filter LPF

LPP :F Data

PIS

LPF :F t---.......J

Block diagram of FQPSK and FQPSK-KF modulator/demodulator (modem). The "F" filter follOwing the Serial to Parallel converter (SIP), . and the CTosscorrelator "X" are described in references [1-7; 13; 17]. Following the crosscorrelator, simple Low Pass Filters (LPF), are used prior to the I and Q modulator. A nonlinearly amplified (NLA) transminer is used.

page 7 K. Feher. UC Davis

September 1993

A

B

" f,.I/T,

o I 0 I 0 O_I~_ A -Il=fijlJ1--f=1 li=:l---1R-...... --f.I--B 0 ! 01 1 0 0,J... ~rl

-+'rt~:Tb I r-L.J

Doc; IEEE P802.11-93/138

___ TIM!

(a)

(b)

(c)

I J' SIGNAL. AntR CROSS CClMILATICN

Fig. 4 Generation of FQPSK and of FQPSK.KF. modulated and NLA radio signals. (a) A basic FQPSK baseband proceSsed SIgnal dt -.tc(JI'SK-t'·, also knOwn as "UF" or Intersymbol and Jitter Free Signal. (b) Filtered ':Fn crosscorrelator concept based on Kato-Feher (KF) invention, abbreviated FQPSK-KF. (c) Baseband I and Q cro~lated sipals. .


September 1993 Doc; IEEE P802.1l-93/138

CLK_ ......... .,..

Fig. 5

3-blt shirt reg.

ROM 512X8

simple digital r 11 ter

A simple Implementation circuit or an FQPSK-KF baseband transm i tter

ROM 512X8

simple digital rllter

uc Davis RA930917

D/A 0-channel output

An illustrative circuit diagram of <a> F-filter. crosscorrelator and seriallparalleVoffset logic [18]. (b) simple digital Low Pass Filter (LPF) implementation [25] based on discrete Ie components. In the N1T paper related to FQPSK [7]. Dr. Kato. et. al.. present several VLSI implementations.

submission page 9 K. Feher. OC Davis


Fig. 6

Submission

1.00

0.00

-1.00

-2.00 0.00 0.50 1.00 1.50

11MB

Hardware - experimentally measured and computer gener~.ted eye diagrams of a subclass of FQPSK and FQPSK-K F' systems. The same type of eye diagrams have been measured on operational radiolwireless products [6].

page 10 K. Feher, UC Davis


in-phase

Fig. 7 Constellation "I-Q" phasor/amplitude diagrams of several unfiltered and filtered FQPSK and FQPSK-KF signals. These diagrams have been measured in linearly amplified systems. In NLA (nonlinearly amplified systems), ideal constant envelopes are obtained. The "crosscorrelator" reduces the envelope fluctuation. prior to the NLA. practically to zero.


September 199 3

"C' 10.0

~

I

-10.0

-20.0

-30.0

-40.0

-50.0

I I

" , , .... , ". , .

" .

FOPSK-l FQPSK-KF A •. 707) FQPSK-KF (A=.4;

,. ,".

1.0

'\ \: \ ~~ .. " "

1.5

\ , \

\

2.0 2.5

PSDofFQPSK

3.0

Doc; IEEE P802.11-931138

3.5 4.0 ITs

Fig. 8 Power Spectral Density (PSD) of a class of FQPSK signals. FQPSK-l is the spectrum of the basic "FQPSK-l" as described in [6] and [13]. Modifications of the basic digital filtered signal, as stated in patents [1-5], lead to a large class of signal shapes and to desirable NLA spectrum and robust BER performance. For FQPSK-KF, a correlation factor of A = 0.707 and a factor of A = 0.4 with a simple LPF processed case is illustrated.



Power(dB) PSD of FQPSM and G~M 8

G

12

18

24

38

3G

42

48 8.8 8.2 8.4

Square : FQPSK Dianond: GnSM BT=&.3 Circle GnSM BT=&.S

8.G 8.8 WTb

Power(dB) PSD of FQPSM and ~X 8

18

28

38

48

58

G8

78

B8 8.8 8.5 1.8

Square : FQPSM DiaMOnd: GftSK BT=&.3 Circle : ~X BT=&.S

1.5 2.& WTb

1.8

2.5""

Fig. 9 The power spectral density advantage in the 20 dB (99%) ranle of the FQPSK as compared to GMSK is illustrated [6). _ .. _ .. a_ .. __ _ ..

submission page 13 K. Feher. UC Davis

September 1993 Doc: IEEE P802,11-931138

Fig. 10 Power Spectral Density (PSD) illustrative nonlinearly amplified (NLA} spectrum at fb = 100 kbls. Several FQPSK type of systeml have been operated in the 16 k~s to S12 k~s ranp.

Submission page 14 K. Fehel'. UC Davis


BE R. (Pe) . . . 01

. . LiM'S" . .. • l-

~ . ':r ' .. I"-Y"'~ " . ~ K ... ·x ...; ....

'" " ' . ·02

"- .... ... ...

" r"..'.'" , ,""" ~" .

", " ~"" ...... '03

~ "' .. ~ " "

'. ~ ~.

i'\ " " OF SK (1r eory), ~ '" -04

,~

-oe 2 3 5 8 7 8

Eb/No (dB). 8 . 10

.. - •. C3MSK 8T-0.5

QPSK (Th.ory).

-+- F -QP8K (H/L) •

•• )C •• C3M8K IT-o.I

11

Fig. 11 BER = f(Ebl'No} perlbnnance of nonlinearly amplified "FQPSK-l" and of coherent GMSK systems in a "stationary" A WON (Additive White Gaussian Noise) environmenL



Fig. 12

Submission

L

I 1.0E-02 L

1.0e-03

1.0e-04 5

. ---- - -~'. -_. - -X : .. . -- ~ .... -

-~ 0 QPal(

_. ' o.

. ~, -_ .. - (I Mal( aT-G.1. . ,-. _._--- " ~, ~, :

~. :

,

"-, ."

: i , I I , I ! 1 I

1

10 15 20 25 30 38

ell (dB). BER = f(C/I)-Rayleigh: Rayleigh faded interference (C/I) controlled perfonnance of coherent FQPSK-1 and of coherent GMSK systems.



Fig. 13

GM81( 8T-G.3.coher ~ f-OP81( (H/L A.p.)

a opal( Clift. Cia.)· - • _. - GMSK noncoherent BTb = 0.3

:..

Pe. -~ . . , '

1.0E-02

>-----~

1.0E-03 r ' ~- ..

~.

1.0E-O" 5

~

:

I 10

' , , "

;

:

15

. ~

,

"" "" " .'

,

" """

, , " ~ . , . :

~~ . , , , . '. , - ::I>..

, I

1 20 25

Eb/No (dB).

, I

I - , -.

.-

""" :

' . i ~

'-. ""

,

" ' -'

" ........

30 35

Rayleigh faded perfOtmance of hardlimited (HIL) or conventional C-class nonlinearly amplified (NLA) FQPSK-l and of GMSK (BTb = 0.3) coherent and noncoherent systems. Note the 5.5 dB advantage of the FQPSK systems when compared to OMSK noncoherent systems. In theoretical studies it was demonstrated that OMSK systems. in general. are more robust than reduced deviation index GFSK systems.


September 1993

Fig. 14

Submission

1.01-02

1.Qe.03

1.De-04

1.0.0s

Doc; IEEE P802.11-93/138

~ ClI. , , . , , ",

• " 'II ... , '

". " q, ". , \. '

\,

\.

" '101 , , "

'C.

'. ~. ~.

". ,'.

,

\'.

'a

, , , , 'a

\ \ •

.-. .- --. .......• .- - -. o •.•• ··0

1.0.0e ...... ---...... ---10--+----'1---1--..... __ - ...... ---10--1

ffi 1.0.01

=

1.0.02

4 5 6 7 8 9 10 11 12 13 14 15

•

~'" ... ". ". , " ,

" , , , , ,

•......• • ___ e

0--0

, , , ,

EItt'No

DEQPSJC("RearJ) CQPSK(1'heory)

FQPSK-KF

~ , , , '.

1.0.03 .4-_-J,.. __ .L-_-I-__ I--_~-~~-~ 6 8 10

BER performance of a class of nonlinearly amplified FQPSK and FQPSK-KF systems in AWGN (stationary) and in Rayleigh faded mobile environment. The thcoretica1linearly amplified coherent QPSK curves are also illustrated.

page 18 K. Feher, tiC Davis

~ (I)

HPA2 (6 dB lower than HPAl)

** HPAl , HPA2 are operated in saturation mode.

Fia. IS Block diagram of an "FQPSK-4*4u system with two nonlinear amplifiers (NLA). "High Power Amplifier" HPA-2 has 4 times (6 dB) lower power than HPA-l. Details have been described in [13; 1-4].

o

m tTl tTl "'tl

September 1993

Fig. 16

Submission

Doc; IEEE P802.11-93/138

0.0

-10.0

-20.0

-30.0

,ITs.

-40.0 L_---!--~:---~..L..:~;;;-.......... ---;~-0.0 0.5

Power spectral density of nonlinearly amplified (NLA) FQPSK-KF 4.:4 signal having a practical spectral efficiency of 2 bls/Hz.


September 1993 Doc; IEEE P802.11-931l38

Fig. 17

submission

~ ~

8-'0 , .• ~

'0 ~ , 0

0.00

-10.00

-20.00 . i

- 30.00

-40.00

-50.00

---- ....

.. ------_.16QAM

A a 1.0 ~~-A-O.7

A-O.g AaO.&

-80.00 L ........ --!.1-..-f,..----L-....... -=-=:--....... 7!::;'-;7:"~ 0.00 1.00 4.00 If - felT.

0.25 0.75 1.00 (f - 'elTb

Normalized frequency

Integrated out-of-band power and eye diagram of a subset of FQPSK-4*. radio NLA applications.



... g .. '0 ~ ~

'" -'l

~

Fig. 18

submission

10-2

to-3

10-·

10-5

Theoretical curve for

10-8 16 OAM in linear channel

10-7

FQPSK-KF(4*4)

~ Rx filters are fOurth.order Butterworth LPFs with BT. = 1.07

Conventional 16 OAM. operating at 9 dB HPA input backoff (x/sin x amplitude-equalizer. ct = 0.4 raised·cosine filters used)

10-8~~~ __ ~~~~ __ ~~~ __ L-~~~ 6.00 10.00 14.00 18.00 22.00 26.00 30.00

Eb/NQ (dB)

BER = f(EblNo) curves for a subset of "FQPSK 4·4" nonlinearly amplified systems.

page 22 K. Feher. Dc Davis

September 1993

w

Coa desiled sipal POWCl'

AO-I = 1st adjacau cbIoneJ inlaferax:e power

S(f)

/ Desired signal power

/lSlACI

III ACI after ItA filLer (our dcfllliliGa)

Doc; IEEE P802. 11-931138

, F!g. 19 Integrated ACI definition in 1st ACI witll ."ric:k war aDd wiOI pndIaI , , budp .. or equivalent lowpw ~I~n. We _ tile ..... ,.. pradlml .....


f'-I

~

"0 ~

~

(a) (b) (c) (d) (e) (f) (g) (h) (i) (j) 1 Modulation 4-FSK 4-FSK GMSK GFSK FQPSK-l FQPSK-KF FQPSK-4x.4 FLOQAM 2 Organization and/or California PrOltim Motorola Motorola UC Davis UC Davis UC Davis Lamair

individual Originator Miaowave '!1 MacDonald MacDonald Fdler Fdler Fdler Chayal/ Silberman [24] [14] [9] [1 to 6] [1 to 7; [1 to 7; Roihenberg

Reference Number (.I'eaently 7-23] 7-23] [24] Wirelesl-l

lIS} 3 In FCC-l MHz Bit

Rate for CJ9tI, 1.4 MbJs 1.92Mbt's 1.0(-) MbJs 1.0 MbJs 1.1 MbJs 1.5 MbJs 2.048+Mb/s 1.72 Mb/s Conservative -20 dB-PSD in [Mb/s)

4 Required Ebl'No for BER = to-Sin [dB] 17 dB 22.5 dB 18.5 dB 19.3 or 15.5 10.5 dB 16 dB 15 dB 16 dB

dB 5 Dep'adation from

linear unplified <-9.9> QPSK-A WON theory <-7.6> <-13.1> <-9.1> or <-1.1> <- 6.3> <-5.6> <6.6> (at 10-5 EtJ <-6.1>

No = 9.4 dB Ref) 6 DeJl'adaIion in

Rayleip mel C/I- 11 TI TI TI <-0.5> <-3> <31> 'l? inJerlerence I dB]

7 CdH::rc:nrIDiffereruiaJ. DlSCR (COH~R) DlSCR DISCR or DlSCR COH COH COH COH

COH 8 other

9 other

10 Key ParIlllCten 4-1evell 4-1evell 2-1evels 2-1evell 2-1evels 2-1evels 4-level not disclosed I

Table 1 Nonlinearly Amplified (NLA) basic modulation/radio parameters of several proposed techniques. Preliminary estimat..:d data based on the availablelindicated references. Draft-l Issue-I; suggested revisions requested from conunittee.


Table 2 Proposed Study for

Data File Transfer Time As a Function of PHY Bit Rate and BER Performance

Single or Diversity?

At "fringe of area/cell?" coverage "geographic" area 80%? 90%?

Assume: Raw BER; no FEe flie length 400 bits to 12,000 bits BER = 10-1; 10-2; 10-4; 10-5

random error distribution

Compare "file transfer time" for a given Ebl'No and CII as function of modem and bit rate. Is it fair to proceed as follows?'

Example: Modem (A) Modem (B)

(1) Ebl'No = 10 dB Ebl'No = 10 dB

BER = 10-2 (?) BER = 10-4

A verage every 101)000 bit Average lout of fue in error. 10 files in error

R-etransmit time ?? Retransmit 10% only

(2) Ebl'No = 15 dB Ebl'No = 15 dB

BER = 10-3 (?) BER = 10-7

Retransmit N times N = 15? Retransmit N = 1 time

File transfer time 15 units Transfer time 1 unit (normalized)

submission page 25 K. Feher, tic Davis

September 1993 Doc; IEEE P8Q2.11-93/138

Bits Required For: Number of Bits (Range) ?? GFSK FQ~SK or other

(noncoherent) (coherent)

1. Tum "on-off' radio transmitter (to reduce 10 10 spectral spreading)

2. Radio receiver and BP filter ringing - other 10 10 transients; AGC?

3. STR (symbol timing recovery or bit timing 20 15 recovery). For coherent joint CR and STR

4. DC offset compensation-discriminator caused 10 0 dc drift due to RF f~equency

5. CR-Carrier recovery synchronization 0 15

6. Switch to higher (or lower) data rates 10 10

7. Message delay (throughput) bits-indicator 5 5

8. OtherTBD 11 11

9. TBD ? ?

10. TBD ? ?

Subtotal 76 76

Table 3 Preambles and PHY headers suggested for study. An illustrative example is shown. Total number of bits might have to be reduced.

Submission page 26 K. Feher. UC Davis


Illustrative eye diagrams for various FQPSK and FQPSK-KF parameters are shown in Fig. 6. Measured eye diagrams on several production units have been the same as the computer generated results. The FQPSK-KF simple processor leads to very easy change of modulation parameters (could be software controlled and changed during operational conditions of WLAN equipment in the field). This is illustrated by several constellation diagrams of Fig. 7.

In Fig. 8 to Fig. 10, several hardlimited NLA power spectral density (PSD) data are shown. In Fig. 11 to Fig. 14, BER curves as a function of EbfNo and interference, i.e., BER = f(CII), of coherently and noncoherently demodulated systems is illustrated. Stationary AWGN (Additive White Gaussian Noise) and mobile Rayleigh faded data are presented.

In Fig. 15 to Fig. 18, block diagram, power spectral density (PSD), and BER perfonnance of several NLA FQPSK 4*4 systems is illustrated.

4. DATA FILE TRANSFER TIMEIMESSAGE DELAY AND THROUGHPUT: A PROPOSED INVESTIGATION

For standardized 1 Mb/s rate GFSK as well as for higher bit rate systems, e.g., 2 Mb/s-FQPSK, we propose to the committee to investigate the "data transfer time" or "message delay time" and throughput and its relation to the robustness of the demodulated radio signal. An oversimplified, however interesting, example is illustrated in Table 2. Geographic coverage "fringe of area/cell" 80%, 90% ...? should be addressed. Initially perhaps the retransmit time or message delay and throughput of two modem/radio systems should be compared, in a nondiversity mode, as indicated in Table 2. Evidently, more realistic and complex scenarios and architectures could be developed later.

From preliminary Table 2 we note that a less robust modem, e.g., modem (A), would require a file transfer time 15 times longer than that of the more robust modem (B). In this illustrative non-diversity case for an EbfNo = 10 dB, we assumed that the less robust modem has a BER = 10-2 while the more robust modem has a BER = 104. Evidently the complete PHY and MAC study group could investigate whether it is really true that a few dB's of C/I or CIN difference in robustness could have a tremendous impact on message delay and on throughput. I believe that in the digital cellular industry 1 dB to 2 dB difference in BER = f(CII) has a tremendous impact on data throughput, capacity, and delay. Is this also true in the WLAN case?

S. SUGGESTED PREAMBLES AND PHY HEADERS

In this suggestion, we address the issue of bit preambles/PHY headers which are required.

(a) Synchronization of the GFSK standard demodulator at 1 Mb/s rate.

(b) "Switch to higher (or lower?) data rates from the data rate defined above in (a).

(c) Synchronization of other than GFSK modems which could require coherent demodulation for higher bit rate systems such as FQPSK, 4-FSK, FLOQAM or others.



We anticipate that the following preamble bits will be required (a very rough estimate of the approximate number of bits is indicated). Number of bits and functions of bits to be studied and modified by the committee.

Frequent switching due to receive diversity switching is assumed. A study of the partial list of illustrative number of bits is recommended, see Table 3. It is assumed that the PRY and MAC committee's will complete, update and modify this table.

In the illustrated example, 76 bits are required for both systems: the standardized noncoherent 1 Mb/s rate GFSlC and for the coherent higher rate modems, e.g., FQPSK or FLOQAM [15; 24]. loint carrier recovery (CR) and symbol timing recovery (STR) design could reduce the overall synchronization time [7]. Note: A "gearup" as well as "gear-down" could be useful. In a poor transmission environment controlled by interference and/or delay sp~ it might be advantageous to reduce the transmission rate from 1 Mb/s to 500 kb/s. A reduced transmission bit rate or PHY rate in such a propagation environment could lead to faster data file transfer as illustrated in Table 2.

In particular for OFSK with variable deviation or modulation index (minimum deviation of 160 kHz was standardized) a lower bit rate could lead to a substantially increased deviation and considerable improvement in the robustness and BER of the system. Reduced (improved) BER could lead to much shorter data file transfer time. The number of preamble and header bits could be too long (or too shon?) for some applications. A study is recommended

6. FCC SUGGESTED CLARIFICATION FOR ''99% POWER" TERM AND REQUEST FOR WIDER THAN 1 MHz FH ASSIGNMENT

The FCC rule in regards to the 99% power in a maximal bandwidth of ±500 kHz has been so far interpreted on a conservative and aggressive manner [9]. I have the impression that the current interpretation of the 802.11 committee is based on the conservative side. It is my understanding that this interpretation requires that the power spectral density or PSD be 20 dB attenuated at 1 MHz (±500 kHz). In Fig. 19, two interpretations based on the integrated ACI (Adjacent Channel Interference) power concept are shown.

Perhaps the committee could raise this issue with the FCC. In several related systems. e.g., IS-54. digital cellular IDMA integrated ACI has been specified. Integrated ACI is (in my opinion) more meaningful because it is the integrated ACI that causes most of the hann. Interpretation based on integrated ACI = -20 dB could lead to substantially increased bit rates [6; 9].

We might also wish to study the potential advantages of obtaining larger than 1 MHz assignment for PH in the 2.4 - 2.48 OHz band.

7. "CREATE" SOFTWARE PROGRAM

The "CREATE" software program has been developed for analysis and design of OMSK, QPSK and a large class ofFQPSK, FQPSK-KX as well as self created-designed FQPSK signals. Offset QPSK (QAM) is the basic structure. Linearly and nonlinearly amplified devices are modeled.

Submission page 28 K. Feher, UCDavis

September 1993 Doc; IEEE P802.11-931138

It is anticipated that this program could become available, free of charge, within several months (probably January 1994). If interested, please write to K. Feher. CREATE has been developed at several universities, under the direction of Dr. Feher. Upgraded, more user~friendly software is currently being developed at the University of California, Davis (UC Davis).

8. SUMMARY

In Table 1 a performance comparison of various modulation NLA techniques is presented. This Table illustrates that the nonlinearly amplified "FQPSK 4·4" technique offers a 2 Mb/s (or even higher rates) in 1 MHz RF bandwidth and for an Ebl'No = 15 dB has a BER = 10~5. It is more robust to noise than any of the listed lower bit rate alternatives, with the exception of FQPSK ~ 1.

The most robust performance in a faded interference and noise controlled environment, is attained by FQPSK~ 1 which requires only an Ebl'No = 10.5 dB for 1.1 Mb/s at BER = 10~5 (as compared to others with at least 15 dB). If binary eye diagrams/simplest decision threshold receivers are of interest, then FQPSK-KF offers a 1.5 Mb/s rate with a robust performance and simple implementation. For further potential bit rate increases clarification, study and interpretation of the FCC rules (integrated ACn and wide bandwidth is suggested.

A detailed "data file transfer time/message delay" investigation is recommended. Initial data indicates that more "robust" (lower EtJNo and C/I requirement) systems could have much shorter delays and increased throughput. Preamble and PHY header bits for the recently standardized 1 Mbls rate GFSK~FH system and for PHY or MAC bilS to negotiate a switch to higher (or lower) data rates are also suggested.


S~tember 1993 Doc; IEEE PS02,11-93/138

9.

[1]

[2]

[3]

[4]

[5]

[6]

[7]

[8]

[9]

[10]

[11]

[12]

[13]

[14]

[15]

[16]

[17]

[18]

REFERENCES

K. Feher. "EiltcL" U.S. Patent No. 4.339.724. Issued July 13. 1982. Canada No. 1130871. August 31, 1982. S. Kalo, K. Feher: "Correlated SiI::nal Processor." U.S. Patent No. 4,567,602. Issued January 28. 1986. Canada No. 1211-517. Issued September 16, 1986. K. Feher: "Modem/radio for nonlinearly amplified systems," patent disclosure files in preparation, Digcom, Inc. Confidential and proprietary, Digcom, Inc., 44685 Country Club Dr., El Macero, CA 95618, December 1992. J. S. Seo, K. Feher: "Superposed Quadrature Modulated Baseband Siena! Processor," U.S. Patent No. 4,644,565, issued February 17, 1987. Canadian Patent No. 1-265-851; issued February 13, 1990. K. Feher: "Notice of Patent Applicability," Document IEEE P .802.11-93/139, Atlanta, September 1993. K. Feher: "FQPSK: A modulation power efficient RF amplification proposal for increased spectral efficiency and capacity GMSK and 1t/4-QPSK compatible PHY standard," Document IEEE P.802.11-93197, Denver, CO, July 13. 1993. S. Kato. S. Kubota, K. Seki. Ta. Sakata. K. Kobayashi. Y. Matsumoto "Implementation architectures. suggested preambles and VLSI components for FQPSK, offset QPSK and GFSK standard 1 Mb/s rate and for higher bit rate WLAN," a submission by NTT-Japan. Document IEEE P.802.11-93/137. J. Socci: "GFSK as a modulation scheme for a frequency hopped PRY." A submission by National Semiconductor. Document IEEE P.802.11-93n6, Denver, CO, July 1993. J. MacDonald: "Discussion of modulation parameters for the 2.4 GHZ FH PRY," a submission by Motorola, Document IEEE P.802.11-93n6. Denver. CO. July 1993. Leung, P.S.K .• K. Feher: "FQPSK - A superior modulation technique for mobile and Personal Communications," IEEE Transactions on Broadcastin~. Vol. 39, No. 2, June 1993. pp. 288-294. Leung. P.S.K.. K. Feher: "F-QPSK - A Superior Modulation for Future Generations of Hi-gh-Capacity Microcellular PCS Systems" Proceedings of the IEEE - VTC-93. May 18-20, 1993, Secaucus. NJ. K. Feher: "Cellular Digital Mobile Radio" manuscript for a forthcoming book Fall 1994 and/or publications. File ©KF. University of California. Davis, September 1992. Feher. K. • Ed.: "Advanced Digital Communications: Systems and Signal Processing TechniQues." Prentice-Hall. Inc., Englewood Cliffs. New Jersey 07632. 1987. J. McDonald, R. DeGroot, C. LaRosa: "Discussion of 0.39 GMSK modulation for frequency hop spread spectrum," a submision by Motorola. Submission to IEEE 802.11 Wireless Access Methods and Physical Layer Specifications, Document IEEE P.802.11 93-97. May 1993. N. Silberman: "Proposal for a modulation technique for frequency hopping spread spectrum PHY standard," a submission by California Microwave, Inc., IEEE P.802.11 93/94, May 10, 1993. Feher, K.: "Modems for Emerging Digital Cellular Mobile Radio Systems," IEEE Transactions on vehicular Technology, Vol. 40, No.2, May 1991. S. Kato, K. Feher: "XPSK: A new cross-correlated phase-shift-keying modulation technique," IEEE Transactions on Communications, May 1983. C. Rueb: "FQPSK carrier recovery/modem design and evaluation," Digital Communications Research Laboratory, Department of Electrical and Computer Engineering. University of California, Davis, CA 95616, September 1993.


StWtember 1993 Doc; IEEE P802.11-93/138

[19]

[20]

[21]

[22]

[23]

[24]

[25]

Y. Guo: "FQPSK-KF additional performance and bit rate improvement alternatives," Digital Communications Research Laboratory, Department of Electrical and Computer Engineering. University of California. Davis, CA 95616, September 1993. H. Yan: "Performance Report/Comparison and Improvements of "20 dB" based PSD criteria for NLA-systems FQPSK and others," Digital Communications Research Laboratory, Department of Electrical and Computer Engineering. University of Califomia, Davis, CA 95616, September 1993. N. Dang: "Performance of a class of 'FQPSK-4-4' nonlinearly amplified radio systems," Digital Communications Research Laboratory. Departtnent of Electrical and Computer Engineering. University of California, Davis, CA 95616, September 1993. H. Mehdi: "Design and simulation of an FQPSK-KF increased spectral efficiency system for FH-spread spectrum based on 4-4 signaling," Digital Communications Research Laboratory, Department of Electrical and Computer Engineering. University of California, Davis. CA 95616. September 1993. R. Atienza: "Baseband processor DSP design of FQPSK-l and FQPSK-KF nonlinearly amplified (NLA) wireless systems," Digital Communications Research Laboratory, Department of Electrical and Computer Engineering, University of California, Davis, CA 95616, September 1993. N. Chayat. M. Rothenberg: "Simulation results for several WLAN modulation methods," a submission by Lannair, Inc. Presented and distributed during the IEEE P.802.11 WLAN High Data Rate FH-PHY Ad-Hoc Group Meeting, Singapore Room, Apple Computer. Cupertino, CA, August 23, 1993. Document IEEE P.802.11-TBD/XX, August 23, 1993. M. Soderstrand, W.Y. Chan: "DSP implementation ofFQPSK-l and FQPSK-KX baseband processors, correlators and filters." Digital Signal Processing and Communications Laboratory, Department of Electrical and Computer Engineering, University of California, Davis, CA 95616. Report MS/WYC, Davis, CA, September 13, 1993.

Submission page 31 K. Feher. UC Davis

1 Mb/s and Higher Data Rate PHYIMAC: GFSK and … 1993 Doc: IEEE P8Q2,ll-931l38 IEEE 101.11 WIRELESS...

Documents

Transcript of 1 Mb/s and Higher Data Rate PHYIMAC: GFSK and … 1993 Doc: IEEE P8Q2,ll-931l38 IEEE 101.11 WIRELESS...